SENSITIVITY AND SPECIFICITY

Doctors order a lot of tests, many of which have a sensitivity and specificity for some condition. On the surface, these terms seem self-explanatory. If a test is sensitive, it's usually positive if a condition is present. If a test is specific, a positive test makes the condition very likely. Most people understand these concepts, but applying the information can sometimes be confusing. To illustrate, I'm going to present you with a clinical scenario.


So what do you want to tell her? Does she really have a 90% chance of having colon cancer? Let's break these two measures down and see if we can give her an educated answer.

First, let's tackle sensitivity, defined as the percentage of people with the disease who get a positive result. If a test has high sensitivity, almost everyone with the disease will have a positive test. Cologuard has a sensitivity of 92%, which means only 8% of people with colon cancer will get a negative result (false negatives). Tests with high sensitivity are most useful when negative because it means the disease is very unlikely. The significance of a positive result depends on the specificity.

Specificity is the percentage of people without the disease who get a negative result. Cologuard has a specificity of 90% which means only 10% of people without colon cancer will get a positive result (false positives). A test with high specificity is most useful when positive because it makes the disease likely. The illustrations below help to explain the relationship. [71]

SELECTION BIAS

To illustrate selection bias, let's discuss a study involving pioglitazone (Actos). Pioglitazone is an insulin sensitizer like metformin that lowers blood sugars without increasing the risk of hypoglycemia, making it a popular drug 10 to 20 years ago. However, it does have one nasty side effect: fluid retention. In studies, up to 30% of heart failure patients experienced worsening symptoms while taking it, and after it had been on the market for a while, a contraindication in NYHA class III and IV heart failure was added to the labeling.

In a recently published cohort study, Korean researchers used a medical database to compare the incidence of dementia between diabetics who took pioglitazone and those who did not. After examining data from 91,218 patients over 10 years, they found dementia risk was significantly lower in patients who received pioglitazone (HR 0.84, 95% CI, 0.75-0.95). [55] This study received a fair amount of press, and if you Google "pioglitazone and dementia," pages of results will come up about these findings. So, should everyone over fifty take pioglitazone? I would argue no because these results are likely due to selection bias.

Consider the following: the two most common causes of dementia are Alzheimer's disease and cerebral atherosclerosis, which causes vascular dementia (strokes, cerebral ischemia). The most common cause of heart failure is ischemic heart disease, also caused by atherosclerosis. Taken together, it is easy to see that patients with heart failure are more likely to have atherosclerosis, which also increases their risk of vascular dementia. [57]

Now back to our study. As we discussed earlier, pioglitazone and weak hearts do not mix well, so it is less likely to be prescribed to patients with heart failure. This is selection bias, and it indirectly reduces pioglitazone's association with vascular dementia. In their conclusion, the authors claim that pioglitazone can "suppress dementia in patients with DM," but buried deep in the paper, they admit the possibility of selection bias. And in case you were wondering, a RCT comparing pioglitazone to placebo for dementia prevention has been performed, and it was stopped early due to futility. [56]

In our confounder model, heart failure is the confounder because it reduces the likelihood of the exposure (pioglitazone) and increases the probability of the outcome (dementia).